Nine Big Air Coolers For Intel's Haswell CPUs, Reviewed

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
[The consensus was that stability at 4.5 and 4.6 GHz on air was extremely rare, that most enthusiasts should expect to top out at 4.3 GHz or less using 1.2 V, and that heat is the most likely culprit.]

I have no idea what motherboards you're referring to in this blurb but my experience makes your article seem like absolute drivel. I have the 4770-K with the ASUS MAximus Extreme VI and H100i, no discrete graphics yet. Build myself, install windows and AI Suite III, clicked literally 4 buttons in the 4 way profiler to get OC of 5.2GHZ.

I run Crysis 3 at very high settings, no discrete graphics for a couple of hours, max out the CPU at 90-95% load and the board and PLX dont get above 30C, and the chip doesn't get above 37C.

So I dunno where you get your setups from but overheating and lack of stability, no way.

Maybe if you dunno what your doing and your gear forces you to be a Jedi Master.

Me, I'll happily admit I don't know much about OC but with AI Suite / OC Panel, seems I dont really need too in order to beat out whomever it is you're referring to in these articles.

Truth is Haswell s going to crap if you slot in a crap board. All performance based systems are about the weakest link. So its waste of time talking about coolers, if you're overworking the chip in a myriad of other ways.

Not trying to be an arse, it's just you read so much crap about this stuff from people whom apparently don't even own the gear they are reviewing. These Haswells, kick arse.

GTX780 has been OS since the board and chip arrived, but I've got no problem gaming with the on-chip 4600, till that arrives. The frame rates aren't mint, but they're %$#%^ awesome at 1920*1080 for on-chip graphics.

Way too many nay-sayers in the review community now. I think it's all monkey hear, monkey say. Form your own opinions, based on your own experiences. Instead one person has a crap setup, so has a crap experience, then tells 10 other people about it and suddenly you have 11 people saying such and such is shite.

Nah, not true, its just one person doesn't know what he's doing and 10 other people are just too lazy or dont have the resources to do it properly.
 
Haswells heat issues are an obvious indicator that the "core i" architecture is played out.
1st gen was a transitional chip which was ok but not great. 2nd gen was an excellent chip in every sense of the word. 3rd gen was a marginal improvement over the sandy bridge chips as far a IPC and they ran hotter. 4th gen is a total fail in a desktop environment IMHO (iGPU improvements and lower wattage good for low power environments), barely any IPC improvement over 3rd gen and tons of heat to boot. It's pretty sad that my 150w OC'd 2nd gen FX spec max's out temp at 62c where as a 70~ watt core i7 max's temp at almost 100c... I'm not trying to hate Im just saying it seems as though this rock has been bled.
 
Haswells heat issues are an obvious indicator that the "core i" architecture is played out.
1st gen was a transitional chip which was ok but not great. 2nd gen was an excellent chip in every sense of the word. 3rd gen was a marginal improvement over the sandy bridge chips as far a IPC and they ran hotter. 4th gen is a total fail in a desktop environment IMHO (iGPU improvements and lower wattage good for low power environments), barely any IPC improvement over 3rd gen and tons of heat to boot. It's pretty sad that my 150w OC'd 2nd gen FX spec max's out temp at 62c where as a 70~ watt core i7 max's temp at almost 100c... I'm not trying to hate Im just saying it seems as though this rock has been bled.
 
Zipperhead - And your one case is supposed to outweigh the dozens of others that can be found easily online? First, I have an extremely hard time believing you hit 5.2GHz on air at all, let alone using Asus 4-Way optimization.

I've used 3 Asus boards and 4770ks with the Phanteks cooler. I've used the Z87 Gryphon, Maximus Gene, and Maximus Hero. My highest clock using ASUS 4 way optimization was 4.5GHz. My personal machine only went up to 4.2GHz. I'm sure I could go higher if I tweaked myself.

My point is, your case, if true at all, is not indicitive of typical Haswell performance. Just look around a bit.

On a side note, the Phanteks is freaking awesome. For those worried about noise, perhaps you should check out SPCRs review on the phanteks. Though it is a DB louder than the D14, it's noise charactaristics are more pleasant to the ear. They recommend the phanteks over the noctua.
 
More likely than not is that his "one case" is completely fictitious. Haswell simply won't run those temperatures when loaded, even at stock settings, on air. One might be able to get a bad reading from PC Probe, but Realtemp should set things straight. And the 95% load thing...the test used in this review is much more strenuous than any game.

That doesn't necessarily mean anyone's intentionally being deceptive. Lots of guys who don't know much about temperature monitoring would believe whatever PC Probe tells them, and these processors are so good at thermal throttling that a user might not even notice. And I can even accept the argument that the stress test used for this specific benchmark is unrealistically harsh, just as many reviewers accepted Nvidia's argument that Furmark was unrealistically harsh.

 
It would be interesting to simulate how each unit worked over time, I.E. collecting dust and reducing airflow. I wonder if that is possible.
 


IB is far from guaranteed to reach 4.8 GHz on air, though. Even many SB's topped out at 4.5 GHz on air. People seem to view the older CPUs through rose-tinted glasses.

Normal range for the 4770K seems to be 4.3 - 4.6 GHz depending on how much heat you're willing to accept. 4670K runs a bit cooler so can more easily reach the higher end of that scale (4.5 - 4.6 GHz). If HW is 10% faster, that means a 4.5 - 4.6 GHz Haswell is equal to a 5+ GHz IB and close to a 5.5 GHz SB.

I'm using the Phanteks cooler and at first I thought the temps in this review seemed *really* low. Then I realized they were temps *above ambient" so you can easily add 30C to those temps for the case temp. That means all of the coolers tested run close the the thermal limit of the CPU (lowered from 105C with IB to 100C with HW). My 4770K with HT enabled @ 4.3 GHz/1.214V runs about 85C in Linpack.
 
I wonder how the lower end and similarly priced water coolers ( Corsair H60 , Antec Kuhler 620 , CM Seidon 120M , NZXT Kraken X40 and others ) would have fought with these air coolers. And most importantly , I don't get it , why would Noctua simply drop their best shot ? That Phanteks V/S Noctua comparison if the best bet I have. And No one remembered that Haswell is just another Intel's "Much Improved lineup compared to older precious gen CPU's in X apps" ? Its great for Non overclockers , but for someone who plays games at 4.3GHz+ , it sucks. Haswell might be 5-14% faster , but is 3-40% more heat generating when OCed. Also it has a complete crap OC record. Unless you have a Fire extinguisher and Liquid Nitrogen cooling , and are out to break the World record on 3DMark along with 4 GTX Titans or the highest temperature reached on a CPU , OCing on Haswell is useless.
 
For the next article, I would recommend taking a haswell and removing the heatspreader as some modders have done to see if it may be possible to get a better overclock with a high quality TIM. I have heard tales that it helps extreme overclocks, but a test with several coolers would be interesting.

Of course Toms Hardware would not want to recommend anyone do this to their CPU, but it would be a fun experiment nonetheless!
 


I'm not wholly surprised, but that is definitely interesting to know. The Scythe tries to push a lot of CFM through (the thickest tested) heatsink without a lot of oomph. Thermalright's often-refreshed design is more suited for that kind of sense.
 


that noctua looks like it would be very similar to my old Thermalright Ultra-120 Extreme, so the short answer is yes... it would benefit a lot from a 2nd fan.
 
YES a nh-14 would probably perform pretty well in that test... though pretty well is relative. None of those air coolers were amazing effective, proving that none of them are really sufficient for haswell. From the benching that's been done on haswell chips, you really do need something more robust then an air cooler... a high end AIO water cooler, or a DIY water loop are probably required now if you want a haswell overclock over 4.5ghz and don't want to buy a bunch of chips looking for a golden overclocker. (there really are foundries which have huge issues with their haswell chips and temps... especially in SE asia, many of those chips won't go past 4.2ghz on air)
 


Would be a nice experiment, but would have limited real-world applications. Most coolers include the fans in the box, so testing with a different fan is no longer testing the product as the manufacturer intended. I also don't believe that many users swap out their cooler fans anyway, so I don't know that this method would represent typical use either.



As Tom said, Xigmatek was there, and Zalman, Noctua, and Phanteks are pretty big players in the cooling market. The Ashura is from Scythe, who many people consider to have some of the quietest and most efficient fans available.
 
Hmmm, what's with the edit and quot buttons - can't find them all of a sudden.

The above comment is well thought out that Haswell runs so hot that any serious overclocker would need a water cooling system. I would add just one point to this.

The NH-D14 and the Plantek and possibly the Silver Arrow all compete fairly well with water coolers like the H100i. Custom water cooling loops do a better job but cost 4-5 times more than the NH-D14 which can regularly be bought on sale for about $75. Say that you got a 4.7 ghz overclock with the custom loop and a 4.4 ghz overclock with the NH-D14 - would you really see much difference in performance?

Add to that all of the extra work that a custom loop takes to install and maintain and if the custom loop leaks, you can quickly trash your computer. Water loop systems can fail.

With the NH-D14 and the other coolers I mentioned, there is no maintenance, no risk of water damage and it is only 3-5 degrees Centigrade behing the custom water loops. It also doesn't cost $300-$400 and is very easy to install. It seems to me to be a no brainer to get an air cooler unless your building a very high-end system.
'
The air coolers also direct heat from other components out of the computer case - water loops do not cool the mother board, the ram and other items in the computer.

My NH-D14 keeps my 2600K CPU below 40C. If I overclocked, I could easily go to 4.6 ghz or 4.8 ghz but I simply don't need the extra speed for what I am doing. Haswell can't go into these ranges without expensive, complicated water cooling.

In the big picture, Haswell is a huge disappointment. It really doesn't impress in speed or efficiency.
 
Read his first statement again, looking for keywords that prompted that specific response. Then look everywhere else for the label you said I applied to everyone else. Upon not finding that label used elsewhere, hang your head in shame.

Anyone who wants an honest debate must practice honesty. We're all a little prone to exaggeration, but the honest thing to do upon realizing a mistake is to fix it.

Edit: Removed
 


Most of this I agree with, and these are largely the reasons I don't bother with water cooling: the marginally better cooling with liquid is not worth the increased expense or time to me. That said, I think you're missing out on something here ( that a lot of other people don't note either. )

A lot of people complain that Haswell ( and IB ) doesn't overclock near as well as SB. In sheer clock frequency, no, the newer Intels don't go as high. However, the IPC efficiency improvement in IB and Haswell usually makes up for the lower clock speeds reachable so you do get similar or better overall performance. It takes a SB @ 4.8GHz to beat an IB @ 4.5GHz, and only barely then ( and SB will use a lot more power doing it, too. ) Even then, you'll still be missing the added instruction support that IB and Haswell have added. The downside is that while IB ran on much less power than SB when OC'd, Haswell increased ( though it's still lower than SB. )
 


i don't know... this h100 was a pretty good deal at $50 new. Got it on sale x2, the first sale was because it was returned to corsair to add a compatible intel mount for a 2011 mb (that cut the price down to $75), the 2nd sale was because the box was opened in store and the mounting brackets and screws removed. knocked the price down to $55. I talked with management and they sold it to me for $49.99 after taxes. I then contacted corsair and they sent me all the mounting hardware and screws free of charge.

One of the better bargain purchases ever ~ works flawlessly and is a large improvement on my very nice Thermalright Ultra-120 Extreme (i think once upon a time getting a Athlon 64 x2 6000+ and motherboard for $4 just because a cap was blown was a better deal, replaced the cap, got a top of the line mb and cpu for $4)



I don't think anyone is discounting this... i think the issue is they're asking for more money for both the haswell cpu and z87 mbs are more expensive as well, and because of the heat issues you get pretty much zero NET improvement over SB or IB. seriously, look at the benches we're talking about 1 or 2 fps difference in pretty much every game. What's the point of improving the IPC if it nets you nothing in practical experience?

 
I have the NH-D14. All you whiners whining why it wasn't included go start your own tech website. Tom's has fully explained the reasons why it was not included.
Click to expand...

Yeah we got it the first time which you would know if you had actually read our posts, we just think the reasons given to leave out the what is probably the highest rated air cooler on the market are not very intelligent.

To clue you in, that is not whining, it is called "critical thinking " - you know all that stuff that science is based on. You should try to learn it some time before you die.
 
sometimes I really wonder: how can people continue to read tom's, and other tech review sites, without realizing that they can ONLY REVIEW what manufacturers send them... no tech site buys their own hardware and just goes ahead and reviews it. if you want those kinds of comparisons, go read forums at overclocker.net or something...
 
warmon6 writes:
> Acording to Linus from LinusTechTips, there suppose to be some black fans coming to market sometime this year. :)

Finally! 8)


SR-71 Blackbird writes:
> My Phanteks Setup:

Blimey, those Scythe fans look meeeean. 😀

I bought the black Phanteks, but just used the default white fans; since they have
black blades, it still looks pretty decent I reckon (pics taken before I stuffed the
case with 580s).



ingtar33 writes:
> My Antec Ninehundred and Two isn't built for the h100 either. ...

I should clarify: my case is designed to hold a water cooler, but because I
have the PSU at the top (see above link; it's to allow for 4 GPUs on the mbd), an
H100 can't go where it would normally be positioned. I suppose in theory I could
mount a WC at the bottom of the case (even though that area is currently used for
intake fans) but do any kits have piping to the block that long so it could reach
the CPU socket way up above? I suspect I'd have to build something custom, but atm
that's beyond my knowledge.



Crashman writes:
> ... If anyone expected progress in the cooling market, they were mistaken.

Didn't toms do an article about this last year? Comparing the latest coolers to the
tried & trusted models of old like the TRUE?

I'm building an oc'd i5 760 system for a friend atm. A used TRUE (with 2 fans!) cost
only 10 UKP off eBay. I'll lap the sucker anyway (lounge coffee table is a permanent
lapping station), but even just as-is the TRUE still performs very well on newer CPUs,
eg. a mere 3 minutes in the BIOS to get a 2500K @ 4.7 stable with excellent temps (on
a Gigabyte GA-Z68XP-UD4, which only cost 45 used). I bag used TRUEs when I can, must
have almost 30 by now, plus three Venomous-X units, two TRUE Blacks, and even a pre-
owned (but unused) Phanteks which cost only 45 (perhaps because it was the red model,
but for benchmarking, who cares?).


iam2thecrowe writes:
> can the noctua be disqualified for having the ugliest fan i have ever seen. It
> really isn't hard to make plastic an appealing colour.....

😀

Must confess I did cave in when looking for a replacement for the rear 140mm fan in
my HAF case, bought the Noctua A15 (same colour styling); looks horrid, but it
does perform very well.


roadrunner343 writes:
> On a side note, the Phanteks is freaking awesome. ...

When the Phanteks was new, I would have been perfectly happy to buy a D14 instead,
but the price difference was huge (D14 was a lot more), though at first the
Phanteks Black was hard to find (the Blue/Red models cost much less). Now it seems
to have evened out a lot; the D14 is more reasonable, but today the Phanteks Black
and the D14 are about the same here (UK), or if styling doesn't matter so much then
the Phanteks Blue is 5 less.



JimmiG writes:
> My 4770K with HT enabled @ 4.3 GHz/1.214V runs about 85C in Linpack. ...

Ouch, for 4.3 that sounds kinda hot. My 2700K @ 5.0 stays under 80C (1.424V in CPU-Z).

Btw, how good are the RAM controllers with HW? My 2700K is running 32GB @ 2133 just
fine. Can Z87 boards with HW cope with max RAM at high speed ok? I don't think I've seen
any site cover this yet.


gunbust3r writes:
> The Phanteks pulls in the best result for CPU temp over ambient yet gets zero
> mention in the results page or wrap up?

Indeed, that surprised me aswell.


flong777 writes:
> The NH-D14 and the Plantek and possibly the Silver Arrow all compete fairly
> well with water coolers like the H100i. ...

Last year I talked to a company in Germany that specialises in water cooling.
I asked them what kind of budget one should have in order to obtain a water cooling
setup which would be usefully better than a top-end air cooler; the guy said
somewhere around 250 UKP. That's quite a jump over an H100i or the good air coolers
we're discussing here. Sometimes I think an H100 (or whatever) is useful more because
it makes it easier to deal with multiple GPUs (the Phanteks is a bif of a squeeze on
some boards) and gives one greater flexibility in setting something up to custom
cool the mbd/chipset. I've used an H100 for a new i7 870 build, but so far I can't
say it's really any better than a Venomous-X with two Scythe fans.


RedJaron writes:
> The downside is that while IB ran on much less power than SB when OC'd, Haswell
> increased ( though it's still lower than SB. ) ...

IMO the real down side is the ludicrous pricing of the newer IB/HW chips, they're
too expensive. Unless one is doing video encoding or something that gains from HT,
I still think a used 2500K, Z68 board, TRUE, etc. is very hard to beat compared to
spending what is now a ruddy fortune on an all-new build for very little gain over
what SB offered (and used SB still offers). The new instructions benefit very few
users and PCIe 3.0 still isn't that much of an issue for achieving good GPU speed.

Also, in many cases the IPC improvements you're referring to were not as good
as you're making out, SB still performed well. It's why Intel crippled IB in the
1st place, ie. soooo many people buying a 2500K and running it at or close to 5.0,
hence no incentive to upgrade. They made the 2500K too good.

If we had real consumer choice, one should still be able to buy the 2500K, but it's
gone. If it was available at a correspondingly lower price because newer products
exist, I guarantee you it'd be a hot seller. If you think about it, the 'range' of
CPUs released by Intel for a particular socket these days is incredibly narrow.
Consider S775: everything from a yucky Celeron 450 to a QX9770, it's a vast range
of chips with a huge performance spread. Now consider a typical S1155 board, the
range is much smaller, and the performance spread is narrower aswell. We've had so
many socket and chipset changes since X58 and for what? Just 15% better from 2700K
to 4770K - as the toms' review title says, "yawn". Where is the modern X58 with lots
of PCIe lanes? It's not X79 since that's very much focused on 6-core users.

How's this for provocative: 2500K = Win7, 3570K = Win8, 4570K = Win 3.1. 😀

I don't mean that literally of course, just how the whole business feels like to me
in terms of the way desktop CPUs and chipsets have been (barely) evolving since
the 2500K came out. I'm not sure which annoys me more: the fact that the latest
CPUs provide so few PCIe lanes, or that consumers are willing to put up with it
when spending large amounts of money on a quad-core "enthusiast" build. Are
people here really willing to say that a Z87/4770K makes them feel as good about
the tech they're suddenly getting their hands on as it did when they built their first
X58/920 rig? (or in my case, P55/870) I very much doubt it. Mid-range and top-end
PC CPU/chipset tech is stagnating, while at the same time prices seem to be rising.

Btw, vmem's sig is a perfect example of what I mean: 2500K/4.7, P67 with a 7970
(same mbd I have). I bet it runs really well. That kind of system is more than enough
to cope with at least another 2 more generations of GPU changes, and even then he
has room to expand CPU power if need be (2700K, etc.)

Ian.

 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom