NH-D15 or smaller cooler??

[Yeltzin]

Hi.

Im going to build a computer with an i7 6700K cpu, and overclock it just a bit..
When asking in a local forum about my choice of components i said i wanted to use the Noctua NH-D15 cpu-cooler, and
that my priority is to have a quiet but fast computer...
But i was told that the NH-D15 is "large as a loaf of bread" and that it is overkill if im not going to overclock to the extreme,
and that the NH-U12s is enough...

However, will the cpu not be cooler with a larger cpu-cooler, and the fans dont have to run as fast??
Dont i gain anything from using the NH-D15 compared to a smaller like NH-U12s??

note : the case will be the Fractal Design Define R5

..thx!!

[quest_for_silence]

http://www.silentpcreview.com/article1462-page6.html

[CA_Steve]

There's more than just size affecting the cooler's performance. There's also the cooler design, how well the cooler's base interfaces with the CPU's IHS, the fan performance, the airflow through the case...

Here's the latest SPCR cooler review chart for a comparison. The relatively small single fanned Scythe Kotetsu holds it's own vs the two fanned D15 (two separate charts - compare the numbers).

If you just plan to overclock a little and not overvolt, the Kotetsu would do fine on the R5. If you plan to overvolt, then a cooler with more area and two fans can be useful.

[QUIET!]

A big tower cooler is great in theory. If it fits and the fan speed is right you will get low temps and low noise.

Along with that its also big and heavy. I like a small computer so I usually can't use the really big heatsinks but if you have the space its still a few pounds hanging off the motherboard, trying to bend it in to an S shape. If your motherboard is horizontal, its not so bad but I don't think the advantages out weigh the drawbacks unless you are OCing the tail off your CPU.

For those reasons I like to see two things in a CPU cooler. First I want to see bolt through mounting with a back plate. Push pins suck and they bend the board more than just the heatsink weight. Second I want to see a thin tower with fairly wide fin spacing and a 120mm fan. The farther from the heatpipe (and CPU), the less effective the heat sink fin area is so thick isn't terribly efficient. Wide fin spacing makes less air resistance so your fan won't be making a high static pressure. 120mm fans are very well developed for silence and air flow because they are so common and 120mm is enough to cool a pretty hot CPU with low RPM.

The best deal in CPU coolers I have got is a $20 Xigmatek Gaia a few years ago. If that fits your case, you can do a lot worse. There are better but you usually pay more.

[Yeltzin]

Thanks for the answers!
You have made me read alot of reviews and articles, and i now plan to use Noctua NH-U14S instead of a NH-D15 .. 935 grams instead of 1320 grams.. Not as big.. Very quiet... Almost as cool as NH-D15.. No problems with tall memories...

We all have our favorites, or like some brands... And i feel safe with buying Noctua fans/coolers..

Thanks again!

[Ken5d]

I thought I should add to this thread rather than start a new one, because my situation is fairly similar to the OP.

I'm going to be building a new i7-6700K computer soon (I've already started gathering the parts piece by piece as they go on sale ). I do plan to overclock it, lightly, to around 4500 or 4600 GHz. And my case will be the same Fractal Design Define R5 as in the OP.

The ambient temps in the Summer here tend to run into the upper 80Fs (low 30sC), so the "cool" air coming in to the case won't be quite as cool as it is for most people.

I've been reading the review articles here and elsewhere for days now, but just can't decide what would be best for me. I do have a concern about those kilogram monsters (plus fan weight) pulling on the motherboard. Though I've had a Prolimatech Megahalems + 2 fans hanging off my current system's motherboard for 6 years, and so far, no problems. (It's < 1kg, though, at "only" 790g without fans.)

Would anyone care to guess whether I could get by with a Scythe Kotetsu? Would I be better off with a Mugen Max? Or should I step all the way up to a Ninja 4 or even a Thermalright Silver Arrow? Or something else ...?

Thanks for your thoughts,

Ken

[quest_for_silence]

Thanks for your thoughts

I think there's not a definitive answer (particularly with the notorious Haswell TIM issue).

This summer the ambient temp was far higher than usual here (up tp 37-38°C, and above 30°C from eight o'clock to almost midnight). I suffered of CPUs high temp just when stress testing the processor with Linpack, never in daily use: my only real problem was mobo temps, and one of my mobos failed with these temps (though the system was a 24/7 one with reduced airflow for quietness sake).

[Ken5d]

37-38C indoors?! I thought I had it bad .... That's a good reminder to me to make sure I do a good job with airflow through the R5.

I haven't been paying attention for several years now, so I'm not aware of the "notorious" Haswell TIM issue. Does it also seem to be an issue with Skylake as well? Or perhaps do you mean that most of the recent cooler tests have been run on Haswell CPUs, and so the results aren't always a true reflection on what a different processor might experience?

[quest_for_silence]

I haven't been paying attention for several years now, so I'm not aware of the "notorious" Haswell TIM issue.

Shrinking of manufacturing process leaded to a problem related to the thermal interface material between the die and the CPU heatspreader. The one used up to Sandy Bridge tended to crack on a rather smaller surface (I don't have a source for this info, at the moment), so Intel changed the material with a more suitable one, which unfortunately had vastly inferior performance. This issue affect Ivy Bridge, Haswell, Haswell-refresh (Devil's Canyon), Broadwell and of course Skylake, and it's more and more evident if you overclock the CPU (though, to be fair, Intel changed it since Devil's Canyon, in order to mitigate the thermal issue).

The only effective solution found is to replace that TIM material with special parts (there's also an article on SPCR), voiding the Intel warranty, but in my experience it's not that necessary, unless you overclock the CPU (I have a couple of Core i5 4690, and they didn't suffered of these exceptionally high ambient temp in normal usage, though they run fairly hotter than I would have desired at low fan spinning, just under 50°C at idle/low loads, over 65°C when gaming, I'd say about 10°C more than usual).
By the way, first delidding tests on Skylake gives a thermal improvement of about 15-20°C less (for the delidded CPU).

[Bearmann]

I thought I should add to this thread rather than start a new one, because my situation is fairly similar to the OP.

I'm going to be building a new i7-6700K computer soon (I've already started gathering the parts piece by piece as they go on sale ). I do plan to overclock it, lightly, to around 4500 or 4600 GHz. And my case will be the same Fractal Design Define R5 as in the OP.

The ambient temps in the Summer here tend to run into the upper 80Fs (low 30sC), so the "cool" air coming in to the case won't be quite as cool as it is for most people.

I've been reading the review articles here and elsewhere for days now, but just can't decide what would be best for me. I do have a concern about those kilogram monsters (plus fan weight) pulling on the motherboard. Though I've had a Prolimatech Megahalems + 2 fans hanging off my current system's motherboard for 6 years, and so far, no problems. (It's < 1kg, though, at "only" 790g without fans.)

Would anyone care to guess whether I could get by with a Scythe Kotetsu? Would I be better off with a Mugen Max? Or should I step all the way up to a Ninja 4 or even a Thermalright Silver Arrow? Or something else ...?

Thanks for your thoughts,

Ken

Me three

Also looking at an i7 6700 build in R5 case. Also plan to pick up the parts as they go on sale when possible. Thinking about the Scythe's Mugen Max, Ninja 4, Kotetsu, or the Noctua U14s. You also have to think about which motherboard you are using and how the heatsink will fit on it. Some heatsinks will interfere with tall RAM. On some motherboards, the U14s blocks the first PCIe slot. It's very nice that Noctua lists each motherboard and which of their coolers are compatible with it. Unfortunately, Scythe doesn't do the same.

[Abula]

As steve already posted, there is more about coolers than size, there are deminishing returns into how much heat can be transfered, its not all about the size and the amount of fans you can mount.

I bought a kotetsu for a friend and cool their 4690 fine, i didnt get much time to play with the setup nor i directly compare it to others coolers i own, but i can see its a very capable cooler for intels sandy/ivy/haswell and probably for skylake. What i can tell you is that i have compared coolers over time, i used to run the HR02 on many setups, then when the HR22 came out i got it, and the temp difference with equal scenario was very small, maybe a degree or two at the most, same happen with the Silver Arrow, there are gains but very small, that dont warrents the high price, cooler like the Scythe Kotetsu and Thermalright HR02 Macho offer a very good cooler with decent fans at a great price. What i do would like to try down the road is a CPU with a bigger plate like the E editions, maybe coolers like NH-D15 or Silver Arrow can perform better with higher TDP and bigger contact area.

What i do like a lot about coolers like the HR02 is that they perform very good under usual load with very low rpms, and on extreme load they are with in the top, probably not the best but with in 2 or 3 degrees, which is pretty solid, and what most should consider when buying and investing the hard earned cash. For me its about keeping the cpu under temp tolerance and doing it quietely, then its about the amount of money i need to reach that point, that said for me its more a hobby and enjoy a lot building and testing, but all builds i do for friends either have an HR02 or Kotetsu.

[quest_for_silence]

all builds i do for friends either have an HR02 or Kotetsu.

You may add the Thermalright TrueSpirit 140 Power, mate: almost as good as an Archon but slightly smaller, in my experience better than the original HR-02 and than the Kotetsu (thermal-wise, though by a small margin), with a very quiet stock fan (the TY-147), all coming with an affordable price tag here (something in between the Kotetsu and the Ninja 4, I think in North America it is even more affordable).

I hope Lawrence Lee will review it, but in case he won't, trust me: it's all good.

[Ken5d]

Luca, thank you for the info on the TIM issue. That's disappointing to learn. I'm very reluctant to delid the CPU, as you probably can imagine. However, perhaps this means I need to err on the side of a more robust cooler.

Barry, thanks, I know to be careful about compatibility with the motherboard, but didn't know that Noctua lists compatible motherboards. The U14s doesn't seem to be quite as effective as the Scythes we're interested in, but it is hopefully good enough. If they can guarantee compatibility that's a big plus. I haven't yet decided on which motherboard to get, though it's almost certainly going to be one of the Asus models.

Abula, I've paid most attention to the tests here on SPCR, and what you wrote is certainly in agreement with the latest test results. However, while their tests are great for comparing coolers relative to each other, obviously they do not necessarily translate to how the coolers would perform within my closed case (to be) with my components. But it is the best we have to go on, and hopefully would "translate".

My priorities are to keep the cpu as cool as possible (where cooler = longer lasting, hopefully) under extremely quiet conditions. I'm willing to overspend a bit if it will give me piece of mind (however included in that piece of mind is that the cooler not be overly heavy). Budget is not a concern compared to cool & quiet. By the way, my wife's system has an HR01 (with vent), and it has performed superbly for 8.5 years now!

I will add the Thermalright Truespirt 140 Power to my short list. I had read a review yesterday (somewhere), and it sounded good, but not finding it on the SPCR list I dismissed it. Thanks again, Luca.

Ken

[CA_Steve]

All of the i7-6700K reviews with overclocking I've seen (but one) used a liquid cooler. So, there isn't a heck of a lot of data yet on temps with air coolers. The only data point I have is a brief mention at Techspot:

Keep in mind the ambient room temperature was just 21 degrees and we were using the Noctua NH-D15.

Under full load the 4.8GHz overclock did see the CPU reach 90 degrees though it remained stable.

Alas, no CPU-Z screenshot or mention of core voltage at 4.8GHz. Plus, all of these damn reviews are with engineering samples rather than purchased parts. So, don't know if there was any cherry picking by Intel marketing. In any case, reading a bunch of reviews, it looks like the various ES parts run stock speed with 1.2 - 1.3V to the core and OCing to 4.6Ghz means raising the core voltage up to 1.3 - 1.4.

QUIET! just built a 6700K system in a somewhat restricted case and used a Thermalright Macho 90 for the cooler. He saw ~40C rise at stock speeds with Prime 95. Maybe he'll chime in and do some OC tests for some temp compares. If not, here's a swag:

- Best case, assume 1.3V needed to go from stock (1.2V and 4.1GHz) to 4.6GHz. Power is proportional to (V^2)*F. So, power increase is about 32%. Assuming cooler thermal transfer is in linear region, it implies temp rise of 53C...ish. (actually, best case assumes stock voltage is 1.3 and oc is 1.4 for smaller percentage increase...but end result is still around 52C).

- Worst case, assume 1.4 needed to go from stock (1.2V and 4.1GHz) to 4.6GHz. Power is proportional to (V^2)*F. So, power increase is about 53%. Assuming cooler thermal transfer is in linear region, it implies temp rise of 61C...ish.

The Macho 90 is a pretty decent cooler for it's size (getting back to it isn't all about cooler mass or fan size). That said, you might see a 5 to 10C improvement in temps moving to a larger cooler for this level of heat dissipation.

So, Kotetsu would be fine for overclock w/ stock voltage. Maybe the Ninja for overvolting.

[Abula]

with a very quiet stock fan (the TY-147)

I agree with you, i have had TY140 from the original HR02 (none Macho), a couple of TY141 from two HR02 Macho, and another two TY141 from my Silver Arrow, but the TY147 that i had from the Macho RevB it was the best, it even went lower on PWM control, if i remember correctly it droped close to 500rpms.

You may add the Thermalright TrueSpirit 140 Power, mate: almost as good as an Archon but slightly smaller, in my experience better than the original HR-02 and than the Kotetsu (thermal-wise, though by a small margin)

Alright, ill add the to friends list =). Btw which of the two is the one you suggest?
Thermalright True Spirit 140 Power
Thermalright True Spirit 140 BW Rev. A

I kinda wished they were $45, i feel the HR02 Macho Rev. B is at the same price level.

I hope Lawrence Lee will review it, but in case he won't, trust me: it's all good.

Hope SPCR does get one to see how it measures to the now king Kotetsu, specially im interested on them reviewing the TY147 black/white.

[Ken5d]

Thank you, Steve; that's very helpful.

I'll go with your (slightly worse) best case scenario, as I wouldn't be likely to overvolt beyond 1.3v. From what I've read, unless I get very unlucky, I shouldn't have a problem getting to 4.5Ghz somewhere between stock voltage and 1.3v.

It actually seems like a lot of trouble to go through for about a 10% increase in cpu speed. Skylake is disappointing in that regard. Maybe I should just be happy enough with whatever I can get with stock voltage overclocking and go with a Kotetsu.

Abula, it's your first link. The second link is a different cooler (not a "Power", which has longer, more visible heatpipes). It does appear to be appealing, though it's considerably heavier than a Kotetsu.

[Abula]

Abula, it's your first link. The second link is a different cooler (not a "Power", which has longer, more visible heatpipes). It does appear to be appealing, though it's considerably heavier than a Kotetsu.

You probably right, the thing is with Amazon you never know, specially with the HR02, there was the original, the Rev A, the Rev B, the Macho, and even some have losen their HR02 tags, like the latest that comes with the black top plate.... someone from nangaminggear needs to place the titles as thermalright has them fully, else becomes confusing and people will miss the latest iterations.

But i agree, seems the power has thicker heatpipes but one thing that i dont like is the 171.5mm height, not all cases can clear that, while the rev A has 165mm, the HR22 has 159mm height which is in line 160mm that most cases place as their limits, that said cases like Fractal can clear like 180 or more, but in cases like my Silverstone FT05 (states 160mm restriction) can bearly fit an Thermalright Silver Arrow IBE (163mm height).

[Ken5d]

I'd noticed that height measurement, but hadn't yet done the research to see if it would fit in a Define R5, so thanks for that ("180 or more"). As I never swap components between cases, as long as it fits the R5, that's fine.

Wow, triple fans on the Silver Arrow? Respect!

[quest_for_silence]

Luca, thank you for the info on the TIM issue. That's disappointing to learn. I'm very reluctant to delid the CPU, as you probably can imagine. However, perhaps this means I need to err on the side of a more robust cooler.

Well, not exactly: probably I didn't stressed it enough but, as I said, in my experience unless you overclock the CPU, unlikely you'll find the usual references are not enough to keep CPU temp under control (so no need for a monster cooler or delidding the heatspreader). More probably that not it won't run as cool as a previous Sandy Bridge, but that's all, and with almost no penalties, noise and reliability wise.

Besides, in my experience it's much more important a good case airflow (than a beefed up cooler), particularly at high ambient temps.

The U14s doesn't seem to be quite as effective as the Scythes we're interested in, but it is hopefully good enough.

AFAIK the NH-U14S is as good as a Kotetsu, in real life scenarios (and probably better with higher TDP CPU, like the LGA2011 or the AMD ones). It's just expensive, so money-wise it isn't the best contender.

Abula, I've paid most attention to the tests here on SPCR, and what you wrote is certainly in agreement with the latest test results. However, while their tests are great for comparing coolers relative to each other, obviously they do not necessarily translate to how the coolers would perform within my closed case (to be) with my components.

That's perfectly true: on the other hand, SPCR did an exceptionally good job with their Silent PC Builds Guides series, where you can find lots of information about how some of those coolers perform inside a case in real life scenarios.
For instance, the Kotetsu in an FT05, the Mugen Max in a Prodigy, the Kotetsu in an R5, and so on.
Though you can't use those information as a comparison table, they offer to you a really valuable insight.

By the way, my wife's system has an HR01 (with vent), and it has performed superbly for 8.5 years now!

Yes, it is still a very capable cooler (as well as some old Ninjas).

Btw which of the two is the one you suggest?
Thermalright True Spirit 140 Power
Thermalright True Spirit 140 BW Rev. A

Actually they are both good: the smaller Rev. A is slightly less performing but offer a wider compatibility (a bit shorter and slanted) and the same good fan. They look like basically a cutted version of respectively the Archon SB-E and Archon IB-E (with six heatpipes rather than eight).

I was particularly impressed by the heat resiliency of the Power version: I used an old Intel X3220 borrowed from a friend as a test bed, it draws around 150W at the wall under linpack. I pumped up the vcore until I draw 330W at the wall, so I was pumping more than 250W into the cooler at 34°C ambient, but no way to reach 80°C with the fan around 1100rpm! The Rev. A slided just above in the low 80es.

About pricing, different location and time mainly matter: I bought it locally at 44 euros, while the Kotetsu was 38 and the Ninja 4 was 49. I didn't checked the Machos price at that time, but I guess it's comparable with the Power.
I thought it can be more affordable in North America because Euro is 14% higher than USD, and there isn't a 23% VAT as here (though Nan's seem to sell it at about the MSRP).

[Abula]

One very interesting thing about the Thermalright TRUE Spirit 140 Power is they seem to have a new improved fan,

Including one Thermalright TY-147A Utra-low noise 300~1300RPM PWM-Fan (MAX21 dBA).

Thats a very nice improvement, specially into how low Intel CPUs are on idle. Hope SPCR gets to review it soon.

[Ken5d]

Luca, thanks again for all that information. I think you're right that good case airflow is more important than the exact cpu cooler I choose, and the R5 gamer build guide does give some very good info. (I was slightly disappointed that he didn't investigate the bottom vent fan a bit further, but I understand the lack of time.)

Nan's seems to be the only US seller of the True Spirit 140 Power. It's too bad it isn't more popular, but that is the cooler on the top of my list at present. But I'll keep watching and reading until I'm ready to buy.

[QUIET!]

All of the i7-6700K reviews with overclocking I've seen (but one) used a liquid cooler. So, there isn't a heck of a lot of data yet on temps with air coolers. The only data point I have is a brief mention at Techspot:

Keep in mind the ambient room temperature was just 21 degrees and we were using the Noctua NH-D15.

Under full load the 4.8GHz overclock did see the CPU reach 90 degrees though it remained stable.

Alas, no CPU-Z screenshot or mention of core voltage at 4.8GHz. Plus, all of these damn reviews are with engineering samples rather than purchased parts. So, don't know if there was any cherry picking by Intel marketing. In any case, reading a bunch of reviews, it looks like the various ES parts run stock speed with 1.2 - 1.3V to the core and OCing to 4.6Ghz means raising the core voltage up to 1.3 - 1.4.

QUIET! just built a 6700K system in a somewhat restricted case and used a Thermalright Macho 90 for the cooler. He saw ~40C rise at stock speeds with Prime 95. Maybe he'll chime in and do some OC tests for some temp compares. If not, here's a swag:

- Best case, assume 1.3V needed to go from stock (1.2V and 4.1GHz) to 4.6GHz. Power is proportional to (V^2)*F. So, power increase is about 32%. Assuming cooler thermal transfer is in linear region, it implies temp rise of 53C...ish. (actually, best case assumes stock voltage is 1.3 and oc is 1.4 for smaller percentage increase...but end result is still around 52C).

- Worst case, assume 1.4 needed to go from stock (1.2V and 4.1GHz) to 4.6GHz. Power is proportional to (V^2)*F. So, power increase is about 53%. Assuming cooler thermal transfer is in linear region, it implies temp rise of 61C...ish.

The Macho 90 is a pretty decent cooler for it's size (getting back to it isn't all about cooler mass or fan size). That said, you might see a 5 to 10C improvement in temps moving to a larger cooler for this level of heat dissipation.

So, Kotetsu would be fine for overclock w/ stock voltage. Maybe the Ninja for overvolting.

I don't intend to over clock because from what I understand I would loose speed step and other energy efficiency features which means higher temperatures and more power draw. For how I intend to use my PC, it doesn't make sense.

I'm sure there are some circumstances that would make an over clock desirable but I think most people do it for bragging rights. That is not enough for me.

Any way, if you read some of the Macho 90 reviews, it is not quite in the same class as the big towers with 120 or 140mm fans but its not too far behind. My case isn't helping matters either so you probably could do 10° better with a big cooler and good case air flow.

As for the TIM, some Haswells were bad, Devil's Canyon was an improvement and Skylake seems to be OK. If you get bad TIM, your core temps are going to be bad no matter what heat sink you use, going big won't cure it. I am satisfied with the temps I'm getting so no reason to de-lid. With my temps, I doubt a delid would lower them more than a few degrees. With an over clock and higher voltage, a few degrees might become 5 or more and de-lid starts to look tempting.

I don't know about Intel's process but most are characterized out to 85c so I like to keep temps at the low 70s or less under stress. I feel like that limit makes life easy for the CPU, it should be low enough to avoid thermal throttling. Lower is better but once it does what I want and I feel the temps are reasonable, I don't worry about it much.

My advise is to figure out what will fit and then look for reviews and comparisons and pick a heat sink that performs well in temps and noise. You can give noise performance a lot of consideration if you have space. With an i7-6700k the temps seem to be pretty low unless you over clock or your heat sink is awful.

[Ken5d]

I appreciate your thoughts, QUIET!. Once I have my system built and running, who knows, I may decide not to overclock. For me it definitely hasn't anything to do with bragging rights, just video rendering speed. If it doesn't make a significant enough difference, I won't bother. I've been running my current system overclocked for about 3 years now. I do hate knowing all the power I'm wasting, but it does save me time, at times.

The Macho 90 definitely does seem to do very well for its size. But having the space in an R5, I might as well go for something a bit bigger. And I'll definitely try to maximize case airflow, as long as it remains silent, or nearly so. But I don't think I'd go as far as de-lidding. Heck, I never even changed the thermal paste on my current system (between the cpu & cooler, obviously), because I didn't want to deal with getting a good finish after removing the cooler's old paste. (Not to mention the old adage, if it ain't broke ....) No way I'd want to try fooling around with a $350 CPU. I'd probably give up overclocking (or certainly over-volting) if that was the only other choice.

[QUIET!]

I intend to boost performance with fast ram and NVMe SSDs. I figure that should do as much as over clocking with only a modest increase in power consumption and heat (compared to slow ram and SATA SSDs). The price is high but good, cheap, fast: pick two.

[Ken5d]

All the real-world benchmarks I've seen show that faster ram doesn't make enough difference to justify the higher price (1% = 3%): http://www.anandtech.com/show/8959/ddr4 ... -crucial/4

I'm not as clear regarding NVMe SSDs. I thought I've seen some (real world) benchmarks which showed only about a 10% or so boost over something like the Samsung 850 Pro, whereas others show a huge increase (double or better). I'm hoping by the time they become readily available, it will be clear to me, and I can get one for not too large a premium.

[Bearmann]

All the real-world benchmarks I've seen show that faster ram doesn't make enough difference to justify the higher price (1% = 3%): http://www.anandtech.com/show/8959/ddr4 ... -crucial/4

I'm not as clear regarding NVMe SSDs. I thought I've seen some (real world) benchmarks which showed only about a 10% or so boost over something like the Samsung 850 Pro, whereas others show a huge increase (double or better). I'm hoping by the time they become readily available, it will be clear to me, and I can get one for not too large a premium.

Well, I think more testing needs to be done with Skylake. My interest is in Photoshop, and here Haswell showed up to to a 12% improvement using faster DDR3 RAM. 12% doesn't seem like much, yet it's probably a greater improvement than moving from Haswell to Skylake.

http://www.xbitlabs.com/articles/memory ... dr3_6.html

Some also suggest that faster RAM will benefit those who are using the integrated graphics processor of the Intel chips, rather than a separate video card. I haven't seen any testing to prove this, however.

As far as NVMe SSD's, I don't think they don't do much for random IOP's, which is what 95% of most users are concerned with. They do boost sequential writes, so if you are working with video or huge databases, then you may benefit. I'm going with typical SSD's in my new build, at least until the price and stability (i.e. technical issues) are equal to standard SSD's.

[Ken5d]

Thanks for that link, Barry. Clearly if starting at 1333, a higher memory speed can make a definite difference. However, DDR4 starts at 2133, and 2400s if not 2666s are easily available at around the same prices. So I was thinking more along the lines of 2400 going to 3200 just doesn't seem worthwhile. Personally, I'll probably be getting 2400s, unless there are some 2666s for about the same price. I'm more interested in being able to get 2x16gb modules, which are not at all easy to come by (or worth their current premium).

That's a good point. I'm mainly interested in speeding up After Effects video processing. It probably involves a lot of sequential writes, however even there I don't think NVMe SSDs are likely to be worth their premium because my main source and destination drives have to be very large. I'd end up copying files to and from SSDs, which still might be faster, but is too much trouble for me. I'll probably stick with Samsung 850 Pro SSDs for my Windows & temp drives. It won't be optimal, but probably good enough for my work.

[Irrelevant]

12% doesn't seem like much, yet it's probably a greater improvement than moving from Haswell to Skylake.

About the same, actually. The Haswell-Skylake difference is usually placed around 10%, on average, but that depends a great deal on the application. In some benchmarks, Skylake's actually a teensy bit slower, and in some it's about 25% faster.

Regardless, 12% is a large performance improvement, period, and positively huge for a change in RAM. For most applications, it takes a massive difference in RAM speed to make a noticeable difference in performance.

Also, FYI, DDR4 is significantly slower per-clock than DDR3 (ie, if it existed, DDR4-1333 would be a lot slower than DDR3-1333). Having said that, any DDR4 you could buy will be roughly as fast as high-ish performance DDR3.

Some also suggest that faster RAM will benefit those who are using the integrated graphics processor of the Intel chips, rather than a separate video card. I haven't seen any testing to prove this, however.

That would make sense, but only for applications that use hardware acceleration for a significant portion of their processing load, and most professional software doesn't, at the moment. Even some 3D modelers are CPU-only.

[CA_Steve]

Some also suggest that faster RAM will benefit those who are using the integrated graphics processor of the Intel chips, rather than a separate video card. I haven't seen any testing to prove this, however.

That would make sense, but only for applications that use hardware acceleration for a significant portion of their processing load, and most professional software doesn't, at the moment. Even some 3D modelers are CPU-only.

Games using the igp is the typical case.